The agent always talks to a *Conn (never a *Server directly).
Replace the Runner interface with the concrete type throughout:
- Runtime.ToolServer is now *toolsrv.Conn
- BuildRuntime takes *toolsrv.Conn
- session.Config, SessionInfra, InfraConfig use *toolsrv.Conn
- All inline type assertions in agent.go deleted — methods called
directly on *Conn (SetEnv, SetCWD, Close, Detach, ListDetachedRaw,
SignalDetached, GetDetachedOutput, DismissDetached, SetAllowTools,
SetOnToolsChanged, Ping, ToolRegistryRevision)
- Added Conn.ToolRegistryRevision() that returns 0 (remote connections
don't support revision tracking; use push-based OnToolsChanged instead)
- Deleted the Runner interface and its test
- Renamed LoadToolOnRunner → LoadToolOnConn
The entire classifier indirection (interfaces, type assertions, RPC
methods for classification, closure wrappers in BuildRuntime) is replaced
by a simple map[string]ToolInfo on Runtime.
The .meta file is read once into ToolInfo. The loop reads fields directly:
rt.ToolMeta[name].ReadOnly (was: rt.CanParallelize(name))
rt.ToolMeta[name].Tier (was: rt.MemoryTier(name, args))
rt.ToolMeta[name].OutputFormat (was: rt.OutputFormat(name))
rt.ToolMeta[name].ResetsCounter (was: rt.ResetsCounter(name))
Removed: ParallelClassifier, MemoryTierClassifier, OutputFormatClassifier
interfaces, toolClassifier type, and all closure-wrapping in BuildRuntime.
The RPC methods on Server/Conn stay (used by ollie-remote) but the agent
no longer routes through them.
run() signature is now run(rt *Runtime, ctx TurnCtx, state state).
The ctx parameter carries both the cancellation context and per-turn
closures — no more separate ctx + tc parameters.
emit() and retryCountdown() also take TurnCtx directly.
- Kill agentConfig struct. run() takes (*Runtime, TurnCtx, state) directly.
TurnCtx holds only 7 per-turn closures; stable config reads come from Runtime.
- Lazy tool refresh: toolsNeedRefresh() checks ToolRegistryRevision() on the
tool server. Skips ListTools IPC when registry hasn't changed. Remote Conn
falls back to always-refetch (same as before).
- AgentConfig.GenParams() method replaces the 16-line manual field copy in
BuildRuntime.
- Remove CfgBackend/CfgModel from Runtime. Callers read cfg.Backend/cfg.Model
directly since they already have the AgentConfig in scope.
The system prompt tool listing now refreshes immediately on tool_load
instead of only on file-watcher events. buildToolListing helper added
to format the listing in the same format BuildPreamble uses.
Also updates kde submodule: GUI auto-switches to newly created agent,
removes remote field from agent dialog.
- Extract hardcoded tool sections into prompts.ToolSections() function
- Replace template system with simple Go code for better maintainability
- Fix refreshToolListing to preserve tool documentation and discovery sections
- Add comprehensive unit tests for prompts and agent packages
- Remove unnecessary tool_sections.md template file
The refactoring simplifies the preamble construction while maintaining
all functionality. Tool listing updates dynamically via refreshToolListing,
while tool documentation requires agent reload (full BuildPreamble call).
Add ollie9p_open, ollie9p_read_fid, ollie9p_close_fid for streaming
file reads. This allows the KDE GUI to read chat/statewait/eventwait
files natively without spawning subprocesses.
Switch from 'read -e' to plain 'read' so multi-line paste can be
captured. Readline was consuming pasted lines individually.
Increase drain timeout to 100ms.
- Add 'o log [-n N]' to read last N lines from agent log (default 1000)
- Rename 'o tail' to 'o chatstream' for clarity
- Update TUI to show log history before streaming chat
Standardize all system messages to use xml-like tags matching the
pattern used by user interruptions:
- <system-consecutive-errors>
- <system-repeated-error>
- <system-replan-required>
- <system-stall-detected>
- <system-task-state-review>
- <system-step-budget-warning>
- <system-step-budget-stop>
Nudge the agent at 50% of step budget to encourage taking action
rather than looping on research. The warning explains that action
tools (file_write, file_edit, shell) reset the counter.
Tools like file_write, file_edit, and shell now reset the agent's step
counter when called successfully. This allows the agent to continue
working without hitting the soft step-budget guardrail as long as it's
making active progress (writing files, running commands) rather than
looping on research/reading.
Changes:
- Add ResetsCounter bool to ToolInfo and MetaFile structs
- Parse resetsCounter from .meta JSON files
- Reset step counter in agent loop when ResetsCounter tool succeeds
- Add resetsCounter: true to file_write, file_edit, shell
- Reduce maxSteps from 50 to 25 (resetsCounter makes this safe)
FormatEvent now takes primitive string arguments instead of agent.Event.
This fixes the dependency direction - format/ should be a leaf package
that agent/, session/, and fs/ depend on, not the other way around.
- Move FormatEvent to format/ package for reuse by both live handler
and replay
- Move LoadConfig to agent/ package where it belongs
- Remove log validator (was noisy, not essential)
- Update replay to use FormatEvent for consistent output format
- Update kde submodule with GUI streaming fix
cmdRead now opens the file and reads incrementally, writing each chunk
to stdout immediately. This fixes streaming for stream-mode files (chat,
statewait, eventwait) which block per-Tread until data arrives.
The previous io.ReadAll approach buffered internally and never flushed
to stdout because stream files never return EOF.
No separate 'tail' command needed — 'read' just works for both
one-shot files (server returns data then EOF) and stream files
(server blocks between chunks).
The event handler was emitting [[[end]]] to close an assistant block
without first closing an open streaming fence (code block). This left
the log in a state where a [[[call:...]]] block appeared nested inside
an unclosed assistant block, triggering a log structure violation.
Now the call and default cases close streamingFence (like the tool case
already did) before writing the [[[end]]] marker.
- Extract all [[[block]]] formatting constants/helpers into format/format.go
(EndTag, RoleDelim, ToolDelim, CallDelim, BlockHeader, ParseBlockHeader, IsEndMarker)
- Replace switch/case ctl dispatchers in fs/handlers.go with generic
ctlDispatch() from new fs/ctl.go
- Add LogValidator state machine to AgentLog enforcing:
- Every [[[block]]] MUST have an [[[end]]] marker
- Blocks SHALL NOT be nested
- Streaming line buffering across partial writes
- Fatal failure on first violation with line-number reporting
- UnclosedAtTeardown() for session/agent teardown checks
- Fix kde/kate ollie_ghost compile: stripPrefixEcho signature mismatch
- Removed from fs/spec.go: prompt.prev, state, offset, context, tail
- Removed handlers: readAgentPromptPrev, readAgentState, readAgentOffset, readAgentContext, readAgentTail
- statewait still exists but now has an inline read handler
- 5 files out of the namespace; 18 agent-level files remaining
- Reduces each agent dir by 5 entries
- HandlerCtx now embeds context.Context for per-operation cancellation/deadline
- Reverted fsedsl signatures back to simple (ctx Ctx) since context is embedded
- requestGenerate handler uses ctx instead of context.Background()
- cmdInject propagates its context parameter instead of using Background
- Delete agent/complete.go (110 lines) and its requestComplete handler
- Remove /complete 9P file from fs/spec.go
- Update KDE kate plugin to build the completion prompt inline and call
ollie-9p rdwr generate instead of ollie-9p rdwr complete
- StripCompletionNoise and stripPrefixEcho moved into C++ helpers
Ollie Agent
Levi Neely